framework
IntelligenceIn Design• system
Minerva
Local-first analytics and personalization engine using vector-like aspects for privacy-preserving insights and retrieval.
analytics
personalization
privacy
local-first
vectors
Minerva is CREO's analytics and personalization engine, designed with a local-first and privacy-centric philosophy. Instead of centralizing raw analytics data, Minerva performs aggregation and analysis primarily on the client through Praetor.
Insights are encoded as abstract aspects that can be composed into vector-like representations for personalization, discovery, and relevance scoring. These representations can be intentionally opaque, preventing reverse engineering of sensitive user data.
Minerva enables intelligent behavior without surveillance, aligning personalization with user autonomy and consent.
Key Highlights
- Local-first architecture with client component
- Praetor exposes a local HTTP service for apps to submit analytics
- User-controlled collection with privacy options
- Opaque vector/aspect encoding (e.g., hashing inputs before encoding)